2 * Copyright (c) 2010, 2011, 2012, 2013 Nicira, Inc.
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at:
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
19 #include "netdev-vport.h"
23 #include <sys/socket.h>
25 #include <sys/ioctl.h>
27 #include "byte-order.h"
34 #include "netdev-provider.h"
37 #include "route-table.h"
39 #include "socket-util.h"
42 VLOG_DEFINE_THIS_MODULE(netdev_vport);
44 /* Default to the OTV port, per the VXLAN IETF draft. */
45 #define VXLAN_DST_PORT 8472
47 #define DEFAULT_TTL 64
49 struct netdev_dev_vport {
50 struct netdev_dev netdev_dev;
51 unsigned int change_seq;
52 uint8_t etheraddr[ETH_ADDR_LEN];
53 struct netdev_stats stats;
56 struct netdev_tunnel_config tnl_cfg;
63 const char *dpif_port;
64 struct netdev_class netdev_class;
67 static int netdev_vport_create(const struct netdev_class *, const char *,
68 struct netdev_dev **);
69 static int get_patch_config(struct netdev_dev *, struct smap *args);
70 static int get_tunnel_config(struct netdev_dev *, struct smap *args);
71 static void netdev_vport_poll_notify(struct netdev_dev_vport *);
74 is_vport_class(const struct netdev_class *class)
76 return class->create == netdev_vport_create;
79 static const struct vport_class *
80 vport_class_cast(const struct netdev_class *class)
82 ovs_assert(is_vport_class(class));
83 return CONTAINER_OF(class, struct vport_class, netdev_class);
86 static struct netdev_dev_vport *
87 netdev_dev_vport_cast(const struct netdev_dev *netdev_dev)
89 ovs_assert(is_vport_class(netdev_dev_get_class(netdev_dev)));
90 return CONTAINER_OF(netdev_dev, struct netdev_dev_vport, netdev_dev);
93 static struct netdev_dev_vport *
94 netdev_vport_get_dev(const struct netdev *netdev)
96 return netdev_dev_vport_cast(netdev_get_dev(netdev));
99 static const struct netdev_tunnel_config *
100 get_netdev_tunnel_config(const struct netdev_dev *netdev_dev)
102 return &netdev_dev_vport_cast(netdev_dev)->tnl_cfg;
106 netdev_vport_is_patch(const struct netdev *netdev)
108 const struct netdev_dev *dev = netdev_get_dev(netdev);
109 const struct netdev_class *class = netdev_dev_get_class(dev);
111 return class->get_config == get_patch_config;
115 netdev_vport_is_vxlan(const struct netdev *netdev)
117 const struct netdev_dev *dev = netdev_get_dev(netdev);
118 const struct netdev_class *class = netdev_dev_get_class(dev);
119 const char *type = netdev_dev_get_type(dev);
121 return (class->get_config == get_tunnel_config
122 && !strcmp("vxlan", type));
126 netdev_vport_get_dpif_port(const struct netdev *netdev)
128 const struct netdev_dev *dev = netdev_get_dev(netdev);
129 const struct netdev_class *class = netdev_dev_get_class(dev);
130 const char *dpif_port;
132 if (netdev_vport_is_vxlan(netdev)) {
133 const struct netdev_dev_vport *vport = netdev_vport_get_dev(netdev);
134 const char *type = netdev_dev_get_type(dev);
135 static char dpif_port_vxlan[IFNAMSIZ];
138 * Note: IFNAMSIZ is 16 bytes long. The maximum length of a VXLAN
139 * port name below is 15 bytes. Still, assert here on the size of
140 * strlen(type) in case that changes in the future.
142 ovs_assert(strlen(type) + 10 < IFNAMSIZ);
143 snprintf(dpif_port_vxlan, IFNAMSIZ, "%s_sys_%d", type,
144 ntohs(vport->tnl_cfg.dst_port));
145 return dpif_port_vxlan;
147 dpif_port = (is_vport_class(class)
148 ? vport_class_cast(class)->dpif_port
152 return dpif_port ? dpif_port : netdev_get_name(netdev);
156 netdev_vport_create(const struct netdev_class *netdev_class, const char *name,
157 struct netdev_dev **netdev_devp)
159 struct netdev_dev_vport *dev;
161 dev = xzalloc(sizeof *dev);
162 netdev_dev_init(&dev->netdev_dev, name, netdev_class);
164 eth_addr_random(dev->etheraddr);
166 *netdev_devp = &dev->netdev_dev;
167 route_table_register();
173 netdev_vport_destroy(struct netdev_dev *netdev_dev_)
175 struct netdev_dev_vport *netdev_dev = netdev_dev_vport_cast(netdev_dev_);
177 route_table_unregister();
178 free(netdev_dev->peer);
183 netdev_vport_open(struct netdev_dev *netdev_dev, struct netdev **netdevp)
185 *netdevp = xmalloc(sizeof **netdevp);
186 netdev_init(*netdevp, netdev_dev);
191 netdev_vport_close(struct netdev *netdev)
197 netdev_vport_set_etheraddr(struct netdev *netdev,
198 const uint8_t mac[ETH_ADDR_LEN])
200 struct netdev_dev_vport *dev = netdev_vport_get_dev(netdev);
201 memcpy(dev->etheraddr, mac, ETH_ADDR_LEN);
202 netdev_vport_poll_notify(dev);
207 netdev_vport_get_etheraddr(const struct netdev *netdev,
208 uint8_t mac[ETH_ADDR_LEN])
210 memcpy(mac, netdev_vport_get_dev(netdev)->etheraddr, ETH_ADDR_LEN);
215 tunnel_get_status(const struct netdev *netdev, struct smap *smap)
217 static char iface[IFNAMSIZ];
220 route = netdev_vport_get_dev(netdev)->tnl_cfg.ip_dst;
221 if (route_table_get_name(route, iface)) {
222 struct netdev *egress_netdev;
224 smap_add(smap, "tunnel_egress_iface", iface);
226 if (!netdev_open(iface, "system", &egress_netdev)) {
227 smap_add(smap, "tunnel_egress_iface_carrier",
228 netdev_get_carrier(egress_netdev) ? "up" : "down");
229 netdev_close(egress_netdev);
237 netdev_vport_update_flags(struct netdev *netdev OVS_UNUSED,
238 enum netdev_flags off, enum netdev_flags on OVS_UNUSED,
239 enum netdev_flags *old_flagsp)
241 if (off & (NETDEV_UP | NETDEV_PROMISC)) {
245 *old_flagsp = NETDEV_UP | NETDEV_PROMISC;
250 netdev_vport_change_seq(const struct netdev *netdev)
252 return netdev_vport_get_dev(netdev)->change_seq;
256 netdev_vport_run(void)
262 netdev_vport_wait(void)
267 /* Helper functions. */
270 netdev_vport_poll_notify(struct netdev_dev_vport *ndv)
273 if (!ndv->change_seq) {
278 /* Code specific to tunnel types. */
281 parse_key(const struct smap *args, const char *name,
282 bool *present, bool *flow)
289 s = smap_get(args, name);
291 s = smap_get(args, "key");
299 if (!strcmp(s, "flow")) {
303 return htonll(strtoull(s, NULL, 0));
308 set_tunnel_config(struct netdev_dev *dev_, const struct smap *args)
310 struct netdev_dev_vport *dev = netdev_dev_vport_cast(dev_);
311 const char *name = netdev_dev_get_name(dev_);
312 const char *type = netdev_dev_get_type(dev_);
313 bool ipsec_mech_set, needs_dst_port, has_csum;
314 struct netdev_tunnel_config tnl_cfg;
315 struct smap_node *node;
317 has_csum = strstr(type, "gre");
318 ipsec_mech_set = false;
319 memset(&tnl_cfg, 0, sizeof tnl_cfg);
321 needs_dst_port = !strcmp(type, "vxlan");
322 tnl_cfg.ipsec = strstr(type, "ipsec");
323 tnl_cfg.dont_fragment = true;
325 SMAP_FOR_EACH (node, args) {
326 if (!strcmp(node->key, "remote_ip")) {
327 struct in_addr in_addr;
328 if (lookup_ip(node->value, &in_addr)) {
329 VLOG_WARN("%s: bad %s 'remote_ip'", name, type);
331 tnl_cfg.ip_dst = in_addr.s_addr;
333 } else if (!strcmp(node->key, "local_ip")) {
334 struct in_addr in_addr;
335 if (lookup_ip(node->value, &in_addr)) {
336 VLOG_WARN("%s: bad %s 'local_ip'", name, type);
338 tnl_cfg.ip_src = in_addr.s_addr;
340 } else if (!strcmp(node->key, "tos")) {
341 if (!strcmp(node->value, "inherit")) {
342 tnl_cfg.tos_inherit = true;
346 tos = strtol(node->value, &endptr, 0);
347 if (*endptr == '\0' && tos == (tos & IP_DSCP_MASK)) {
350 VLOG_WARN("%s: invalid TOS %s", name, node->value);
353 } else if (!strcmp(node->key, "ttl")) {
354 if (!strcmp(node->value, "inherit")) {
355 tnl_cfg.ttl_inherit = true;
357 tnl_cfg.ttl = atoi(node->value);
359 } else if (!strcmp(node->key, "dst_port") && needs_dst_port) {
360 tnl_cfg.dst_port = htons(atoi(node->value));
361 } else if (!strcmp(node->key, "csum") && has_csum) {
362 if (!strcmp(node->value, "true")) {
365 } else if (!strcmp(node->key, "df_default")) {
366 if (!strcmp(node->value, "false")) {
367 tnl_cfg.dont_fragment = false;
369 } else if (!strcmp(node->key, "peer_cert") && tnl_cfg.ipsec) {
370 if (smap_get(args, "certificate")) {
371 ipsec_mech_set = true;
373 const char *use_ssl_cert;
375 /* If the "use_ssl_cert" is true, then "certificate" and
376 * "private_key" will be pulled from the SSL table. The
377 * use of this option is strongly discouraged, since it
378 * will like be removed when multiple SSL configurations
379 * are supported by OVS.
381 use_ssl_cert = smap_get(args, "use_ssl_cert");
382 if (!use_ssl_cert || strcmp(use_ssl_cert, "true")) {
383 VLOG_ERR("%s: 'peer_cert' requires 'certificate' argument",
387 ipsec_mech_set = true;
389 } else if (!strcmp(node->key, "psk") && tnl_cfg.ipsec) {
390 ipsec_mech_set = true;
391 } else if (tnl_cfg.ipsec
392 && (!strcmp(node->key, "certificate")
393 || !strcmp(node->key, "private_key")
394 || !strcmp(node->key, "use_ssl_cert"))) {
395 /* Ignore options not used by the netdev. */
396 } else if (!strcmp(node->key, "key") ||
397 !strcmp(node->key, "in_key") ||
398 !strcmp(node->key, "out_key")) {
399 /* Handled separately below. */
401 VLOG_WARN("%s: unknown %s argument '%s'", name, type, node->key);
405 /* Add a default destination port for VXLAN if none specified. */
406 if (needs_dst_port && !tnl_cfg.dst_port) {
407 tnl_cfg.dst_port = htons(VXLAN_DST_PORT);
411 static pid_t pid = 0;
413 char *file_name = xasprintf("%s/%s", ovs_rundir(),
414 "ovs-monitor-ipsec.pid");
415 pid = read_pidfile(file_name);
420 VLOG_ERR("%s: IPsec requires the ovs-monitor-ipsec daemon",
425 if (smap_get(args, "peer_cert") && smap_get(args, "psk")) {
426 VLOG_ERR("%s: cannot define both 'peer_cert' and 'psk'", name);
430 if (!ipsec_mech_set) {
431 VLOG_ERR("%s: IPsec requires an 'peer_cert' or psk' argument",
437 if (!tnl_cfg.ip_dst) {
438 VLOG_ERR("%s: %s type requires valid 'remote_ip' argument",
443 if (tnl_cfg.ip_src) {
444 if (ip_is_multicast(tnl_cfg.ip_dst)) {
445 VLOG_WARN("%s: remote_ip is multicast, ignoring local_ip", name);
451 tnl_cfg.ttl = DEFAULT_TTL;
454 tnl_cfg.in_key = parse_key(args, "in_key",
455 &tnl_cfg.in_key_present,
456 &tnl_cfg.in_key_flow);
458 tnl_cfg.out_key = parse_key(args, "out_key",
459 &tnl_cfg.out_key_present,
460 &tnl_cfg.out_key_flow);
462 dev->tnl_cfg = tnl_cfg;
463 netdev_vport_poll_notify(dev);
469 get_tunnel_config(struct netdev_dev *dev, struct smap *args)
471 const struct netdev_tunnel_config *tnl_cfg =
472 &netdev_dev_vport_cast(dev)->tnl_cfg;
474 if (tnl_cfg->ip_dst) {
475 smap_add_format(args, "remote_ip", IP_FMT, IP_ARGS(tnl_cfg->ip_dst));
478 if (tnl_cfg->ip_src) {
479 smap_add_format(args, "local_ip", IP_FMT, IP_ARGS(tnl_cfg->ip_src));
482 if (tnl_cfg->in_key_flow && tnl_cfg->out_key_flow) {
483 smap_add(args, "key", "flow");
484 } else if (tnl_cfg->in_key_present && tnl_cfg->out_key_present
485 && tnl_cfg->in_key == tnl_cfg->out_key) {
486 smap_add_format(args, "key", "%"PRIu64, ntohll(tnl_cfg->in_key));
488 if (tnl_cfg->in_key_flow) {
489 smap_add(args, "in_key", "flow");
490 } else if (tnl_cfg->in_key_present) {
491 smap_add_format(args, "in_key", "%"PRIu64,
492 ntohll(tnl_cfg->in_key));
495 if (tnl_cfg->out_key_flow) {
496 smap_add(args, "out_key", "flow");
497 } else if (tnl_cfg->out_key_present) {
498 smap_add_format(args, "out_key", "%"PRIu64,
499 ntohll(tnl_cfg->out_key));
503 if (tnl_cfg->ttl_inherit) {
504 smap_add(args, "ttl", "inherit");
505 } else if (tnl_cfg->ttl != DEFAULT_TTL) {
506 smap_add_format(args, "ttl", "%"PRIu8, tnl_cfg->ttl);
509 if (tnl_cfg->tos_inherit) {
510 smap_add(args, "tos", "inherit");
511 } else if (tnl_cfg->tos) {
512 smap_add_format(args, "tos", "0x%x", tnl_cfg->tos);
515 if (tnl_cfg->dst_port) {
516 uint16_t dst_port = ntohs(tnl_cfg->dst_port);
517 if (dst_port != VXLAN_DST_PORT) {
518 smap_add_format(args, "dst_port", "%d", dst_port);
523 smap_add(args, "csum", "true");
526 if (!tnl_cfg->dont_fragment) {
527 smap_add(args, "df_default", "false");
533 /* Code specific to patch ports. */
536 netdev_vport_patch_peer(const struct netdev *netdev)
538 return netdev_vport_is_patch(netdev)
539 ? netdev_vport_get_dev(netdev)->peer
544 netdev_vport_inc_rx(const struct netdev *netdev,
545 const struct dpif_flow_stats *stats)
547 if (is_vport_class(netdev_dev_get_class(netdev_get_dev(netdev)))) {
548 struct netdev_dev_vport *dev = netdev_vport_get_dev(netdev);
549 dev->stats.rx_packets += stats->n_packets;
550 dev->stats.rx_bytes += stats->n_bytes;
555 netdev_vport_inc_tx(const struct netdev *netdev,
556 const struct dpif_flow_stats *stats)
558 if (is_vport_class(netdev_dev_get_class(netdev_get_dev(netdev)))) {
559 struct netdev_dev_vport *dev = netdev_vport_get_dev(netdev);
560 dev->stats.tx_packets += stats->n_packets;
561 dev->stats.tx_bytes += stats->n_bytes;
566 get_patch_config(struct netdev_dev *dev_, struct smap *args)
568 struct netdev_dev_vport *dev = netdev_dev_vport_cast(dev_);
571 smap_add(args, "peer", dev->peer);
577 set_patch_config(struct netdev_dev *dev_, const struct smap *args)
579 struct netdev_dev_vport *dev = netdev_dev_vport_cast(dev_);
580 const char *name = netdev_dev_get_name(dev_);
583 peer = smap_get(args, "peer");
585 VLOG_ERR("%s: patch type requires valid 'peer' argument", name);
589 if (smap_count(args) > 1) {
590 VLOG_ERR("%s: patch type takes only a 'peer' argument", name);
594 if (!strcmp(name, peer)) {
595 VLOG_ERR("%s: patch peer must not be self", name);
600 dev->peer = xstrdup(peer);
606 get_stats(const struct netdev *netdev, struct netdev_stats *stats)
608 struct netdev_dev_vport *dev = netdev_vport_get_dev(netdev);
609 memcpy(stats, &dev->stats, sizeof *stats);
613 #define VPORT_FUNCTIONS(GET_CONFIG, SET_CONFIG, \
614 GET_TUNNEL_CONFIG, GET_STATUS) \
619 netdev_vport_create, \
620 netdev_vport_destroy, \
626 netdev_vport_close, \
630 NULL, /* recv_wait */ \
634 NULL, /* send_wait */ \
636 netdev_vport_set_etheraddr, \
637 netdev_vport_get_etheraddr, \
638 NULL, /* get_mtu */ \
639 NULL, /* set_mtu */ \
640 NULL, /* get_ifindex */ \
641 NULL, /* get_carrier */ \
642 NULL, /* get_carrier_resets */ \
643 NULL, /* get_miimon */ \
645 NULL, /* set_stats */ \
647 NULL, /* get_features */ \
648 NULL, /* set_advertisements */ \
650 NULL, /* set_policing */ \
651 NULL, /* get_qos_types */ \
652 NULL, /* get_qos_capabilities */ \
653 NULL, /* get_qos */ \
654 NULL, /* set_qos */ \
655 NULL, /* get_queue */ \
656 NULL, /* set_queue */ \
657 NULL, /* delete_queue */ \
658 NULL, /* get_queue_stats */ \
659 NULL, /* dump_queues */ \
660 NULL, /* dump_queue_stats */ \
662 NULL, /* get_in4 */ \
663 NULL, /* set_in4 */ \
664 NULL, /* get_in6 */ \
665 NULL, /* add_router */ \
666 NULL, /* get_next_hop */ \
668 NULL, /* arp_lookup */ \
670 netdev_vport_update_flags, \
672 netdev_vport_change_seq
674 #define TUNNEL_CLASS(NAME, DPIF_PORT) \
676 { NAME, VPORT_FUNCTIONS(get_tunnel_config, \
678 get_netdev_tunnel_config, \
679 tunnel_get_status) }}
682 netdev_vport_tunnel_register(void)
684 static const struct vport_class vport_classes[] = {
685 TUNNEL_CLASS("gre", "gre_system"),
686 TUNNEL_CLASS("ipsec_gre", "gre_system"),
687 TUNNEL_CLASS("gre64", "gre64_system"),
688 TUNNEL_CLASS("ipsec_gre64", "gre64_system"),
689 TUNNEL_CLASS("vxlan", "vxlan_system")
694 for (i = 0; i < ARRAY_SIZE(vport_classes); i++) {
695 netdev_register_provider(&vport_classes[i].netdev_class);
700 netdev_vport_patch_register(void)
702 static const struct vport_class patch_class =
704 { "patch", VPORT_FUNCTIONS(get_patch_config,
708 netdev_register_provider(&patch_class.netdev_class);